DX Unified Infrastructure Management

Hi,

UIM 8.47 on windows 2008 R2. We have around 1000 servers to be monitored . As I understand from uim documentation, 100 robots can report to single hub and so on. 

Can we install multiple hubs i.e. 5 hubs in a single high end server? Or 1 hub in 1 server?

Rgds,

Jagdeep

I would not suggest installing multiple robots and hub on one machine.

this is not a tested configuration for hubs.

I would expect you would end up running into performance issues support would not be able to do much about

other than suggest you separate the hubs onto dedicated systems.

As your hubs would have to be using non standard ports for communication this would also increase the complexity of your installs are robots communicate and fail-over to other hubs on port 48002 which will not be the case for 4 of your 5 hubs.

I have plenty of customers running 2000 robots per hub.  Just not on the Primary Hub.

1 hub on 1 server, please.

Notes for Scalability / Capacity.

1. Technical

HUB license key. You can not go beyond total number of Robots per hub which is allowed in HUB license key.

2. Practice.

Probe Data generation interval, number of targets, network bandwidth, etc lots of factors determine the outcome.

Thanks All for your replies.

We have 1000 Robots and have 3 servers for UIM.

1 would be - UIM server

1- would be- UMP

1 would be - Database

All 1000 robots will report to UIM server having primary hub.

Is this correct configuration?

Do not 1000 Robots talk to the primary hub, since primary hub is very busy for own tasks.

So your environment will look like.

1 - Database

2 - UIM Server

3 - UMP Server

4 - Secondary HUB - you should configure your 1000 Robots talking to 4.

Thanks Yu !!

As I have 3 servers only, can secondary hub be installed in one of 3 servers.

i.e. Can secondary hub be installed on UMP server?

Unfortunately no.

Please prepare 4th machine.

Thanks Yu !!

This confusion came because CA support mentioned that per hub will report 100 robots only.

As with all things Nimsoft the answer to your question depends a lot on the details of what you are actually going to be doing with regards to monitoring.

Many of the standard replies depend on assumptions about your host and client OS versions as well as what sort of monitoring you are doing.

You mention "monitoring 1000 servers" and ask if a hub could do that. If you are running net_connect, syslog, and the SNMP listener, then sure, one hub could handle that (unless the volume of data or frequency of tests was too much) if you had a big enough server.

Where it gets tricky to answer the question is when you start using probes with persistent connections. So, suppose you are running the sql_response probe and there are 100 databases on each of those 1000 servers that you need to make independent connections to. You could be running those queries once a day which gives you plenty of time to execute them but if they all have to run at the same time you are going to run out of ports on your hub server.

If those 1000 servers are all robots, then each of those robots will make a persistent connection to your hub. Maybe you'll have enough ports - maybe not - depends on how your OS is configured.

Switching gears, if you are monitoring 1000 clients, you really need to be thinking about redundancy. put some dollars into clustering your database server and get a dedicated secondary to the primary hub. IF you have an infrastructure failure you don't want to be in the position of losing all functionality.

-Garin

Thanks Garin !!

We would only be using cdm + oracle/sql server probe for 100 robots and cdm probe only for remaining 900 robots.

I hope this will not affect performance.

Thanks in advance.